Abstract:<p class="0abstract">Classification and selection of gene in high dimensional microarray data has become a challenging problem in molecular biology and genetics. Penalized Adaptive likelihood method has been employed recently for classification of cancer to address both gene selection consistency and estimation of gene coefficients in high dimensional data simultaneously. Many studies from the literature have proposed the use of ordinary least squares (OLS), maximum likelihood estimation (MLE) and Elasti… Show more
“…In general, Ordinary Least Squares (OLS) is used to solve the regression model (1). That is, to minimize the target function about OLS β :…”
Section: Ordinary Least Squares and Variable Selection Problemmentioning
confidence: 99%
“…With the development of data science and information technology, the problem of high-dimensional data generally exists in biomedicine, machine learning and other fields [1][2]. As regression model has a strong explanatory power on the causal relationship between response variables and influence factors, Regression analysis has always been a popular data analysis method.…”
Regression analysis has always been a popular method of data analysis because of its good interpretation of causality. Variable selection can effectively reduce the complexity of the model and improve the modeling accuracy. As a classical variable selection method, Lasso can realize variable selection and parameter estimation at the same time, but the optimal regularized parameters may lead to inconsistent variable selection results. Adaptive Lasso improves the consistency of variable selection by adopting the least square estimation of coefficients as the weight correction regulars. However, the design matrix of least square estimation is not invertible for ill-conditioned data. In this paper, a modified Adaptive Lasso based on regularization weights and its ADMM regularization algorithm are proposed, taking the Ridge estimation of coefficients as the weight parameters to modify the Lasso regularization terms. Numerical experiments with sparse coefficients, large coefficients, grouping effects, and polynomial regression simulation experiments are carried out to compare the proposed method with Lasso, Adaptive Lasso, Elastic net and the Adaptive Elastic net. The results show that the proposed Adaptive Lasso based on regularized weight has higher prediction accuracy and more sparse variable selection, which effectively improves the accuracy of regression model.
“…In general, Ordinary Least Squares (OLS) is used to solve the regression model (1). That is, to minimize the target function about OLS β :…”
Section: Ordinary Least Squares and Variable Selection Problemmentioning
confidence: 99%
“…With the development of data science and information technology, the problem of high-dimensional data generally exists in biomedicine, machine learning and other fields [1][2]. As regression model has a strong explanatory power on the causal relationship between response variables and influence factors, Regression analysis has always been a popular data analysis method.…”
Regression analysis has always been a popular method of data analysis because of its good interpretation of causality. Variable selection can effectively reduce the complexity of the model and improve the modeling accuracy. As a classical variable selection method, Lasso can realize variable selection and parameter estimation at the same time, but the optimal regularized parameters may lead to inconsistent variable selection results. Adaptive Lasso improves the consistency of variable selection by adopting the least square estimation of coefficients as the weight correction regulars. However, the design matrix of least square estimation is not invertible for ill-conditioned data. In this paper, a modified Adaptive Lasso based on regularization weights and its ADMM regularization algorithm are proposed, taking the Ridge estimation of coefficients as the weight parameters to modify the Lasso regularization terms. Numerical experiments with sparse coefficients, large coefficients, grouping effects, and polynomial regression simulation experiments are carried out to compare the proposed method with Lasso, Adaptive Lasso, Elastic net and the Adaptive Elastic net. The results show that the proposed Adaptive Lasso based on regularized weight has higher prediction accuracy and more sparse variable selection, which effectively improves the accuracy of regression model.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.